Development of CAD system for 3D breast ultrasound images

We have developed a CAD system for masses on 3D breast ultrasound images and estimated the preliminary performance with about 200 cases. The detection and classification methods were based on edge detection and active balloon models to segment the mass regions. Volume rendering technique was also applied to prepare the 3D images to readers. It showed a good classification performance as an initial result. Especially, the classification performance showed a good rate because the cases employed in classifications scheme were difficult ones to discriminate by the ultrasound images in a physician's opinion, however the number of the case we used was not so large at this time. keywords: ultrasound, mass, classification, CAD.